Statistical analysis of the GSM (Global System for Mobile Communications) mobility prediction models

Research Paper (postgraduate), 2017

51 Pages


Table of Content


1.1 Preamble
1.2 Significance of Mobility Prediction

2.1 Introduction
2.2 Obstacle Mobility Model
2.3 Street Unit Model
2.4 City Area Mobility Model
2.5 Random Walk Model
2.6 Random Waypoint Model
2.7 Markovian Random Walk
2.8 Random Direction Model
2.9 Shortest Path Model
2.10 Normal Walk Models
2.11 Smooth Random Mobility Model
2.12 Microscopic Models
2.13 Mesoscopic Models
2.14 Macroscopic Models
2.15 Reference Point Group Mobility
2.16 Column Mobility Model
2.17 Pursue Mobility Model
2.18 Nomadic Community Mobility Model
2.19 Activity-Based Model
2.20 Cell-Residence-Time-Based Model
2.21 Pathway Mobility

3.1 Introduction
3.2 Taxonomy Development
3.3 Taxonomy Parameters

4.1 Introduction
4.2 Models Based On Location and Fixed Velocity
4.3 High Probability Prediction Models
4.4 Cell to Cell Mobility Models
4.5 Models Based On Feasible Future Sequence of Cells
4.6 Models Based On Degree Of Randomness
4.6.1 Trace-Based Models
4.6.2 Constrained Topology Models
4.6.3 Statistical Models
4.7 Models Based On Level of Description
4.7.1 Fluid Flow Model
4.7.2 Gravity Mobility Models
4.8 Models Based On Individual User Behaviors
4.9 Models Based On Nodes Movement Dependency
4.9.1 Temporal Dependency of Velocity
4.9.2 Spatial Dependency of Velocity
4.10 Models Based On Real Network Movement Depiction

5.1 Introduction
5.2 Models Based On Location and Fixed Velocity
5.3 Models Based On High Probability Prediction
5.4 Cell to Cell Mobility Models
5.5 Models Based On Feasible Future Sequence of Cells
5.6 Models Based On Degree of Randomness
5.7 Models Based On Level of Description
5.8 Models Based On Individual User Behaviors
5.9 Models Based On Nodes Movement Dependency
5.10 Models Based On Real Network Movement Depiction

6.1 Introduction
6.2 Requirements for the Proposed Model




In the GSM environment, mobility prediction is concerned with envisaging of the mobile station’s next movement. By accurately employing the predicted movement, the GSM network is capable of attaining enhanced resource allocation and reservations, better assignment of cells to location areas, more efficient paging, and call admission control. Numerous studies have been carried out in mobile station prediction and as such, many mobility predictions have been developed. The goal of this paper was to statistically analyze these mobility prediction models in order to understand their strengths and weaknesses. The results of this analysis indicated that the current mobility models base their predictions on movement patterns of users throughout physical space. However, they do not directly relate to movement throughout the network based on the cell presently hosting a mobile station. In addition, several base transceiver stations may overlap over a single physical location, a complication not captured by the current mobility prediction models. Therefore, there is need for a generic mobility prediction model that can depict movement of mobile users more realistically within the GSM network coverage areas. As such, this paper proposes the development of a mobility prediction model that is capable of accurately reflecting mobile station movement in real-world cellular networks, taking into consideration the actual scenarios such as base transceiver stations overlapping. The requirements for this novel mobility prediction model are also provided, and were based on the shortcomings noted in the existing mobility prediction models. The significance of the proposed model lies on the fact that in order to pre-allocate resources for seamless connectivity during handovers, the mobility prediction model should anticipate the actual cell that the mobile station will next connect to, rather than the physical location they will move to.

Keywords : GSM, Mobile station, Mobility prediction


1.1 Preamble

The emergences of information technology and telecommunication services have had a huge impact in people’s day to day lives. Globally, each and every sector of people’s economic, social and cultural lives has been affected by these technologies. In particular, mobile communication services have been accepted as part of office procedure. An increasing number of organizations are allowing their employees to utilize their mobile stations to carry out office work, a concept referred to as Bring Your Own Device (BYUOD). In the homes, mobile stations are crucial in keeping in touch with family members. Effective mobile communication requires an accurate mobility prediction model so that network services can be delivered to them without long network paging delays.

As Mona et al., (2015) found out, mobile communication services have led to some security concerns. This owes to the increased flexibility that allows criminals to commit crime from one location and move virtually to other locations and still enjoy perfect cellular communications. Terrorists are also increasingly using mobile communication services to coordinate their activities. Any effort towards counter-terrorism or crime prevention requires that there be a more efficient geographical location management scheme so that law enforcement departments can track the culprits with increased precision.

As a result of the huge uptake of mobile communication services, most GSM networks can be easily overwhelmed with traffic data. In highly populated locations such as cities, chances of connection delays, numbers of uncompleted and dropped calls are common scenarios (Abdou et al., 2015). This can be attributed to the problems that are inherent in the present process of updating and searching the current locations of multiple mobile nodes in a GSM network.

The challenges outlined above all stem from the fact that the current mobility prediction models are not effective enough to provide an accurate depiction of the actual mobile user movements within the GSM coverage area. This paper sought to unravel the statistical genesis of these mobility models shortcomings.

1.2 Significance of Mobility Prediction

The ultimate goal of the GSM mobility prediction models is to make an attempt in imitating the movement of real mobile stations, which are characterized by the change of speed and direction with time (Petteri, 2016). This allows the network to track the location where the subscribers are currently residing. In so doing these mobility modes permit voice calls, short messages (SMS), general packet radio services (GPRS) and other mobile phone services to be delivered to the subscribers.

According to Nweke et al., (2015), mobility prediction models are significant in the provision and maintenance of communication with a mobile user at any given point in time. This is particularly true now that there has been a growing trend of the convergence of numerous financial services such as banking applications with mobile communication services. However, with the swift growth in the number of mobile subscribers globally, mobility prediction has emerged as one of the most important and challenging tasks for mobile communication systems.

The mobility management enables the serving network to locate a mobile subscriber’s point of attachment for delivery of data packets, and maintenance of a mobile station’s connection as it continues to change its point of attachment (Nandeppanavar et al., 2010). However, with the amalgamation of data traffic on GSM networks coupled with the increasing demand for improved throughput and security, mobile station movement prediction has become a serious concern.

In their study, Garud et al., (2015) indicated that mobility prediction is an important factor contributing to the overall performance of mobile telecommunication networks. This is due to its ability of restraining the ability of the GSM network in maintaining a connection or guarantying a quality of service among the subscribers. In addition, the ever increasing number of telecommunication customers has radically increased the consequences of poor mobility prediction on network maintenance.

Patle and Sanjay (2016) discuss that due to the significance of mobility prediction, several studies have been conducted and their results have led to the development of a number of mobility prediction models such as random waypoint, random walk, random Markovian walk, random direction mobility, smooth random mobility, cell-residence-time-based, Gauss–Markov, Fluid Flow, normal walk, shortest path, activity-based, pursue mobility, nomadic community mobility, reference point group mobility(in-place mobility, overlap mobility, convention mobility), Manhattan grid, pathway mobility, obstacle mobility, Freeway mobility, Street Unit, Street Pattern Tracing, mobility vector, gravity models, city section mobility, city area mobility, First order Kinetic, Feynman-Verlet, Semi-Hidden Markov, Autoregressive (AR), Global Mobility among others.

The rest of this paper is organized as follows: Part II provides a review of the current GSM mobility prediction models while Part III explains the developed taxonomy for the GSM mobile stations mobility prediction models. Part IV carries out some parametric analysis of the mobility prediction models whereas Part V offers a critic of the models. Lastly, Part VI presents the requirements for an ideal GSM mobility prediction model and the conclusion that can be draw from this study.


2.1 Introduction

According to Chuyen et al., (2014), numerous mobile station prediction models have been developed to assist in the forecasting of mobile stations movement patterns within the network coverage area. The most common ones are discussed in the following sub-sections.

2.2 Obstacle Mobility Model

In this model, mobile station mobility is depicted by taking into consideration real-life scenarios such as the fact that people move towards specific destinations rather than randomly choosing some destinations; obstacles such a buildings, parks or rivers can block people’s movements as well hinder signal propagation; and that people do not walk along random directions but along pathways and select shortest paths (Aarti et al., 2012).

2.3 Street Unit Model

Here, the mobile station is permitted to move on a rectangular, Manhattan grid only, where the grid depicts the street pattern of suburban or urban areas. The mobile station speed is selected from a normal distribution and is updated periodically or can be area dependent. Tarik et al., (2011) discuss that direction changes can occur at every crossroads, where the probabilities can be different for each of the four possible directions at every crossroads. Furthermore, other models can be considered in this point due to their comparable characteristics with Street Unit Model, such as high-way traffic models. These models are able to describe the mobility behaviour of mobile stations with an accuracy of a few meters, and hence are useful for devising efficient and effective dynamic channel assignment algorithms.

2.4 City Area Mobility Model

This model is utilized to describe mobile station mobility and traffic behaviour within a city area environment. The transport theory states that although each individual city area exhibits specific characteristics, they share generic features that can be considered as assumptions for representing mobile station’s movement. The first assumption is that the population density gradually decreases as one shifts towards the city edges in suburban and rural area. Biju et al., (2010) point out that on the contrary, densely populated areas surround the city centre with high density of workplaces and shopping centers. The second conjecture is that the street network supports two movement types, radial and peripheral. The last postulation is that the geographical area covers the whole city area, consisting of a set of zones connected via high capacity routes. A zone is regarded as corresponding to a network area, such as macro cell and streets are regarded as high capacity routes. The stochastic part corresponds to the mobile station mobility behavior such as initial distribution, type of movement, criterion for selecting routes, and traffic behavior exampled by call arrival rates and available services.

2.5 Random Walk Model

This is an individual mobility model that is memory-less since it does not retain knowledge related to its past speed and direction. Consequently, the mobile station future velocity is independent of the current velocity (Patle and Sanjay, 2016).

2.6 Random Waypoint Model

Rogerio and Roberto ( 2016) discuss that this is a simple stochastic model in which a mobile station moves on a restricted continuous plane from its current position to a new location by randomly choosing its destination coordinates, its speed of movement, and the amount of time that it will pause on arriving at the destination.

2.7 Markovian Random Walk

This is a modified form of the random walk model that utilizes Markov chains to describe the mobile station movement, and it introduces memory in the movement behavior of the mobile station (Rong-Hua et al., 2015). Also referred to as the probabilistic version of the random walk model, this model employs three states to represent the movement coordinates x and y: state zero (0) represents the current position of the mobile station; state one (1) represents the previous position of the mobile station; while state two (2) represents the next position of the mobile station.

2.8 Random Direction Model

This model is similar to the Random Waypoint Model, only that instead of choosing a destination, the mobile station randomly selects a direction from a given interval and moves in that direction (Jogendra and Panda, 2016). After some random time taken from an exponential distribution, the user either changes direction or changes speed. The movement can occur freely anywhere in the network coverage area. The values for the direction are taken from a uniform distribution on the interval and the values for the speed follow a uniform distribution or a normal distribution.

2.9 Shortest Path Model

This model is utilized to represent the mobility of vehicular mobile stations. It assumes that within the location area, a mobile station follows the shortest path measured in the number of cells passed through, from source to destination. According to Swati and Hina (2014), at each intersection, the mobile station makes a decision to proceed to any of the neighboring cells such that the shortest distance assumption is maintained.

2.10 Normal Walk Models

This is straight-oriented mobility model referred which assumes that a mobile station moves in unit steps on a Euclidean plane (Hala, 2015). In this model, the next movement direction is chosen from a normal distribution with zero mean.

2.11 Smooth Random Mobility Model

This is an enhanced random mobility model that makes the movement trace of individual mobile stations more realistic than common approaches for random movement (Logambal and Chitra, 2016). In this model, the movement pattern is based on random processes for speed and direction control in which new values are correlated to the previous ones. Upon a speed change incident, a new target speed is selected, and acceleration is set to achieve this target speed. The principles for the direction changes are also similar to those of the speed change.

2.12 Microscopic Models

These models depict the movement of a single mobile station by its space and speed coordinates at a given time . The goal here is to obtain a very detailed representation for one entity within the network coverage area. Such models include Street Unit Models and Street Pattern Tracing Models. In the former, the mobile station is permitted to move on a rectangular grid only. Both the Manhattan and freeway models fall in this group. Joanne (2014) explains that in the latter models, the mobile station is allowed to move on a predefined stretch only. As such, they are good at depicting movements along highways or main streets where directions changes are very unlikely to occur. In these models, only the speed of a mobile station is selected randomly from a uniform or normal distribution. The direction is given by the position of the mobile station within the highway or street. At an intersection of a horizontal and a vertical street, the mobile station can turn left, right or go straight with certain probability. This model is therefore ideal for mimicking the motion pattern of mobile stations on streets defined by maps.

2.13 Mesoscopic Models

These models depict the homogenized movement behaviour of several mobile stations instead of only one. Daniel et al., (2016) point out that here, the mobile users shift as groups (hence group models). Examples of these models include Reference Point Group Mobility (In-Place Mobility, Overlap Mobility, and Convention Mobility) and Mobility Vector model. The goal here is to derive a distribution function of the number of vehicles at a certain location ( ) or speed in order to describe the movement of the group. In these models, each mobile station has a logical centre and the trajectory of the group as a whole is represented by the locus of the centre. Though each mobile station has its own reference point, the group moves as a single entity because the reference points follow the group movement. In-place mobility, a given geographical area is partitioned such that each and every subset of the original area is allocated to a specific group, and each group operates only within their geographic subset. It is therefore ideal for simulating scenarios in which groups of people, that have similar goals, are assigned to restricted areas. For the case of Overlap Mobility, several different groups, each of them having diverse tasks, working in the same geographic area are described. Here, each group may have varying characteristics compared with other groups within the same geographical boundary. In Convention Mobility, both the conference attendees and the revelations are represented. In addition, different revelations are housed in different rooms and the rooms are connected to facilitate travel between exhibits. Moreover, this model partitions a given area into smaller subsets and permits the groups to move in a similar pattern throughout each subset.

2.14 Macroscopic Models

In these models, Fabrício et al., (2015) elaborate, the focus is on density, mean speed, speed variance, and traffic flow of vehicles. Examples of these models include fluid flow models, gravity models and the random walk models.

A. Fluid Flow Model:

This model is employed to investigate the average number of mobile stations crossing a boundary of a network coverage area. Tao et al., (2016) discuss that it derives from transportation theory and describes the movement of a group of users. As such, the Reference Point Group Model is included in this family. It can be employed for both intra-cell and inter-cell movements of mobile stations. It works by averaging the mobility patterns of all mobiles stations and as such, it is often used to describe the aggregate mobility behavior of all mobile stations.

A. Gravity Mobility Models

These models are also derived from transportation theory and they give an aggregated description of the movement of several users, as was the case for the fluid model. They are based on Newton’s gravitational law and spatial interactions such as attractiveness and repulsiveness among the mobile stations (Mariano at al., 2016). In Newton’s gravity model, the movement of a mobile station from a given point to another is directly proportional to the attraction of the area and inversely proportional to the distance of separation between them.

C. Random Walk Models

These models depict the movement of individual mobile stations from cell to cell. According to Adebiyi et al., (2016), a state transition diagram defines this model, in which a cell is represented by a state and the mobile station movement is represented by transition probabilities between the states. As such, these models are interested in the cell where the mobile station resides and not its exact location.

2.15 Reference Point Group Mobility

In this model, each group has a center that can be either a logical center or a group leader mobile station. Here, each group is made up of one leader and a number of members (Jaswant and Rajneesh, 2016). The movement of the group leader influences the mobility behavior of the entire group.

2.16 Column Mobility Model

This is utilized to describe a set of mobile stations such as robots moving in a certain fixed direction. As Ahmad (2016) explains, it can be employed in searching and scanning activity, such as destroying mines by military robots.

2.17 Pursue Mobility Model

This mimics scenarios where several mobile stations endeavor to capture single mobile station ahead. As such, it could be utilized in target tracking and law enforcement. Here, the mobile station being pursued serves as the target mobile station and it moves freely in accordance with the Random Waypoint model (Kim et al., 2017). The pursuer mobile stations direct their velocity towards the position of the targeted mobile station in an attempt to intercept it.

2.18 Nomadic Community Mobility Model

This model serves to symbolize the mobility scenarios where a group of mobile stations travel together and can therefore be applied to depict mobile communication in a conference or military application. Mostafa (2016) expound that in this model, the entire group of mobile stations shifts randomly from one location to another and the reference point of each mobile station is established based on the common movement of this group.

2.19 Activity-Based Model

In activity-based model, instead of using a set of random variables to depict the mobility pattern, it is assumed that a trip is undertaken for participating in an activity such as shopping at a given destination. Martin et al., (2016) illustrate that once the location for the next activity has been determined, the route from the current location to this activity location is computed in terms of cells traversed.

2.20 Cell-Residence-Time-Based Model

The idea here is to establish the connection time spent by a mobile station within one location. According to Chuyen et al., (2014), this requires the tracing of the movement of individual users. In their paper, Yi-Bing et al., (2011) used the standard counter values such as the number of handovers and call traffic measured in a mobile telecommunications network to derive the cell residence times. However, measurement of cell residence times in a commercially operated mobile network is not trivial.

2.21 Pathway Mobility

In this model, geographic constraints are integrated into the mobility model by restricting the mobile station movement to the pathways in the map. In these maps, the vertices represent the buildings of the city, and the edges depict the streets and freeways between these buildings. Mahmoud et al., (2016) discuss that initially, the mobile stations are randomly positioned on the edges of the graph and then for each mobile station, a destination is selected arbitrarily and the mobile station moves towards this destination via the shortest path along the edges. On arrival at the destination, the mobile station pauses for time and again chooses a new destination for the next movement.


Excerpt out of 51 pages


Statistical analysis of the GSM (Global System for Mobile Communications) mobility prediction models
Information Technology Security & Audit
Catalog Number
ISBN (eBook)
ISBN (Book)
File size
1011 KB
statistical, global, system, mobile, communications
Quote paper
Vincent Nyangaresi (Author)Silvance Abeka (Author)Solomon Ogara (Author), 2017, Statistical analysis of the GSM (Global System for Mobile Communications) mobility prediction models, Munich, GRIN Verlag,


  • No comments yet.
Read the ebook
Title: Statistical analysis of the GSM (Global System for Mobile Communications) mobility prediction models

Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free